Revealed: Mark Zuckerberg's dramatic plan to replace thousands of employees with AI
Last January, Meta executives developed a secret initiative, Project OT, aimed at replacing thousands of employees with AI agents. The plan was ultimately abandoned due to technical limitations and internal pushback.
Last January, Mark Zuckerberg and senior Meta executives gathered at his estate in Hawaii. The goal was a dramatic redesign of the technology giant's workforce structure and its adaptation to the AI era.
Under the code name Project OT, the executives crafted an ambitious plan designed to turn Meta and its subsidiaries — Facebook, Instagram, and WhatsApp — into AI-first companies. A new Reuters report reveals the dimensions of that secret initiative and the reasons for its collapse.
Tens of thousands of employees to be replaced by AI
Zuckerberg's vision was revolutionary. Instead of relying on human employees to perform tasks such as development, coding, design, or DevOps, these tasks were to be handed over to AI agents. These "virtual employees" were intended to work under the close supervision of a very small number of staff.
Sources familiar with the plan told Reuters that under extreme scenarios, Meta executives considered cutting the workforce in some teams by an unprecedented 60 percent.
The report reveals that this massive organizational change was planned in two waves: the first in May, and the second in November. The move included direct layoffs, closing open positions, and the quiet removal of low-performing employees (PIP). However, just before implementation, the plan went awry.
The technology was not ready
In May, Meta carried out a round of cuts, laying off about ten percent of its employees, but simultaneously froze future plans for aggressive November cuts.
The Reuters investigation points to several factors for the sudden halt. Engineers and product managers realized that internal AI initiatives were intended to make their roles redundant, leading to an internal rebellion that troubled senior management.
Beyond human resistance, the technology was not ready. The expected productivity improvement from the agents was insufficient: the agents made errors in code, failed to understand complex contexts, and required constant "babysitting," which negated time savings. Thus, while code fixes jumped by 220%, the number of features delivered to users increased by only 36%.
Meanwhile, Wall Street investors began questioning what Meta was gaining from its massive expenditures on hardware and AI chips if operational efficiency was not improving as promised.
Meta's response
In response to the report, Meta confirmed the existence of the project but attempted to soften its implications. The company defined Project OT as a strategic move to reduce costs and redirect employees to priority areas, such as generating data for training large language models. Meta admitted that extreme scenarios included reducing teams by more than half, but categorically denied any concrete intention to fire such a huge number of employees all at once.
